Drop your data as it is · fragmented, multi-rate, oddly named. This runs the real intake: it parses and unifies it, measures your sensor noise, flags faulty hardware, tells you honestly what your data can support, and hands back a clean dataset and a readiness report. Works the same for any domain. Your data is processed server-side in a non-retained session and not stored · and the sufficiency and noise method never ships to your browser.
1 Data in
Paste a CSV (a timestamp column plus your sensor columns · wide format), or load a messy sample. Your domain sets the honest thresholds.
Not ready to upload? Try a sample dataset.
Click one to run it through the Cleanroom and see the honest readiness result end-to-end. Each card says what it shows. Your own data replaces it later.
real = actual IndPenSim penicillin data (CC-BY 4.0, Goldrick 2019). representative = a faithful, illustrative dataset for that domain, not raw logged rows · your real data replaces it.
2 Parsing · confirm what each column is
We auto-detected roles. Fix any that are wrong · state = a measured process variable, input = a setpoint / actuator you control, target = the outcome you care about (often a lab result), run id = which batch/cycle, time = the timestamp. Unrecognized tags are kept and flagged, never dropped.
3 Guidance · what your data honestly supports
Sensor noise & hardware
Measured from your data · noise-dominated sensors flagged.
Your guarantee
4 Data out
A cleaned, unified preview (in your own column names) and two downloads: the cleaned dataset and a readiness report. This is your data · we never keep it.
What happens next. This is the readiness step · it runs in a non-retained session and hands back your cleaned data + this report. In a pilot engagement, the next step is where we ground and certify the twin on this data in a non-retained clean-room and return the frozen twin + proof packet (a certificate you verify offline). The two artifacts here (cleaned CSV, readiness JSON) are your data in your own terms · they carry no GREENBOX method or schema.